envloop-cli 1.28.0__tar.gz → 1.30.0__tar.gz

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: envloop-cli
-Version: 1.28.0
+Version: 1.30.0
 Summary: EnvLoop CLI and Python SDK for jobs, training, checkpoints and inference
 Project-URL: Homepage, https://envloop.ai
 Project-URL: Documentation, https://docs.envloop.ai
@@ -249,7 +249,7 @@ The platform starts the published Revision’s `main.sh` / `main.py` using the R
 CLI output uses compact timestamps and highlights statuses and log levels by default on terminals.
 Use `el --no-color job list` or append `--no-color` to a command to disable highlighting, including errors.
 `NO_COLOR=1`, `TERM=dumb`, pipes, and redirected output also disable colors. JSON preserves original values.
-List times use the local timezone; Job logs retain the recorded timezone, with full precision and offsets available through `--json`.
+List and table times are UTC with a `UTC` label; Job logs retain the recorded timezone and mark lines collected only after the Job ended with `(backfilled)`. Full precision, offsets and the `backfilled` field are available through `--json`.
 `el job run live RUN --follow --json` emits complete response frames as NDJSON; Ctrl-C stops reading without stopping the Run.
 
 ## Job-bound decision plugins
@@ -283,7 +283,7 @@ download a complete registered ANS object directory through EnvHub. Datasets req
 authorization the SDK uses GET URLs valid for up to five minutes, verifies file size/SHA-256,
 and installs the directory atomically. Matching platform support is required.
 
-A compatible server and an explicitly granted project Key with `envloop:data:read` are required; writes additionally need `envloop:data:write`. Default and older Keys do not gain these permissions. The current login selects the base URL, workspace and project; there is no `--env` flag. After configuring the Key using `el auth config`:
+A compatible server is required. Data uses the permission of the resource it touches: read (for example `datasets.read`) to query and download, write to publish and upload, delete to retire; a key from `el auth login` follows your project role. The current login selects the base URL, workspace and project; there is no `--env` flag. After configuring the Key using `el auth config`:
 
 ```bash
 el data list --type dataset --json

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/README.pypi.md

@@ -232,7 +232,7 @@ The platform starts the published Revision’s `main.sh` / `main.py` using the R
 CLI output uses compact timestamps and highlights statuses and log levels by default on terminals.
 Use `el --no-color job list` or append `--no-color` to a command to disable highlighting, including errors.
 `NO_COLOR=1`, `TERM=dumb`, pipes, and redirected output also disable colors. JSON preserves original values.
-List times use the local timezone; Job logs retain the recorded timezone, with full precision and offsets available through `--json`.
+List and table times are UTC with a `UTC` label; Job logs retain the recorded timezone and mark lines collected only after the Job ended with `(backfilled)`. Full precision, offsets and the `backfilled` field are available through `--json`.
 `el job run live RUN --follow --json` emits complete response frames as NDJSON; Ctrl-C stops reading without stopping the Run.
 
 ## Job-bound decision plugins
@@ -266,7 +266,7 @@ download a complete registered ANS object directory through EnvHub. Datasets req
 authorization the SDK uses GET URLs valid for up to five minutes, verifies file size/SHA-256,
 and installs the directory atomically. Matching platform support is required.
 
-A compatible server and an explicitly granted project Key with `envloop:data:read` are required; writes additionally need `envloop:data:write`. Default and older Keys do not gain these permissions. The current login selects the base URL, workspace and project; there is no `--env` flag. After configuring the Key using `el auth config`:
+A compatible server is required. Data uses the permission of the resource it touches: read (for example `datasets.read`) to query and download, write to publish and upload, delete to retire; a key from `el auth login` follows your project role. The current login selects the base URL, workspace and project; there is no `--env` flag. After configuring the Key using `el auth config`:
 
 ```bash
 el data list --type dataset --json

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/_optional.py

@@ -30,7 +30,7 @@ FEATURE_COMMANDS = {
 
 MODULES: dict[str, str] = {
     **dict.fromkeys('data data_publications data_upload data_cli profile_updates recipes'.split(), 'data'),
-    **dict.fromkeys('train train_progress train_runner_cli train_log_cli job_logs job_cost inference decision sandbox runtime_params job_spec_cli training'.split(), 'experiment'),
+    **dict.fromkeys('train train_progress train_metrics_cli train_runner_cli train_log_cli job_logs job_cost inference decision sandbox runtime_params job_spec_cli training'.split(), 'experiment'),
     **dict.fromkeys('agent_runs rsi_cli'.split(), 'rsi-loop'),
 }
 CLIENTS = {

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/auth_cli.py

@@ -20,6 +20,9 @@ def add_auth_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParse
     login.add_argument("--no-browser", action="store_true", help="Print the login link without opening a browser")
     login.add_argument("--timeout-seconds", type=int, choices=range(1, 601), metavar="SECONDS", default=300,
                        help="Wait up to 1–600 seconds (default: 300)")
+    login.add_argument("--read-only", action="store_true",
+                       help="Issue a read-only key, for example for CI that only reads results. "
+                            "Without it the key follows your project role")
     configure = commands.add_parser("config", help="Save Base URL and API key for CLI and SDK")
     configure.add_argument("--no-check", action="store_true", help="Save without API verification (offline; key remains unverified)")
     actions = configure.add_subparsers(dest="config_action")

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/auth_login.py

@@ -76,7 +76,8 @@ def login(args: argparse.Namespace, store: Any) -> int:
     def digest(value: str) -> str:
         return hashlib.sha256(value.encode()).hexdigest()
     response = _request(url, 'start', {'key_hash': digest(key), 'poll_hash': digest(poll_token),
-                                    'code_hash': digest(code)}, min(5, args.timeout_seconds))
+                                    'code_hash': digest(code), **({'read_only': True} if args.read_only else {})},
+                        min(5, args.timeout_seconds))
     identifier, uri = _verification_url(url, response)
     interval, expires = response.get('interval'), response.get('expires_in')
     if (isinstance(interval, bool) or not isinstance(interval, (int, float)) or not 0 < interval <= 10 or
@@ -88,6 +89,8 @@ def login(args: argparse.Namespace, store: Any) -> int:
     try:
         print(f'Base URL: {url}\nConfirmation code: {code}\nOpen: {uri}', flush=True)
         print('Sign in with Clerk, enter this code, and authorize your workspace/project.', flush=True)
+        print('The key will be read only.' if args.read_only else
+              'The key follows your project role. The page lists its permissions before you authorize.', flush=True)
         if not args.no_browser:
             try:
                 opened = webbrowser.open(uri, new=1)

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/cli_errors.py

@@ -27,10 +27,10 @@ def _register(codes: str, message: str) -> None:
         _MESSAGES[code] = message
 
 
-_register("DATA_SCOPE_REQUIRED", "Configure a project key with explicit Data read permission, plus Data write for publication and upload management.")
+_register("DATA_ACCESS_REQUIRED", "This key lacks the permission named above on the Dataset, Runner, Profile, Checkpoint or Trial artifact. Run el auth login to issue a new key that follows your project role: every role can read, and Developer or Admin can also write and delete (create, update, delete, upload, runner register). A key from el auth login --read-only cannot write; log in again without --read-only. Or give a key Read or Edit on that page in the Console under CONFIGURE > API Keys. Existing keys never gain permissions.")
 _register("DATA_NATIVE_INCOMPLETE DATA_NATIVE_UNAVAILABLE", "Native catalog registration is incomplete. Retry the same input after the platform is ready.")
 _register("DATA_DEPENDENCIES_ACTIVE DATA_DEPENDENCIES_UNKNOWN DATA_DEPENDENCIES_CHANGED DATA_DELETE_CONFLICT", "Retirement requires a complete, current dependency check. Resolve active or unknown dependencies, then retry the same resource.")
-_register("DATA_PUBLICATION_INVALID", "Check the immutable ANS URI, native manifest and exact file sizes and SHA-256 values.")
+_register("DATA_PUBLICATION_INVALID", "Check the immutable ANS URI, native manifest and exact file sizes and SHA-256 values. A record Dataset source is https://github.com/OWNER/REPO or https://huggingface.co/datasets/ORG/NAME with the full commit, not a branch or tag.")
 _register("DATA_PIN_CONFLICT DATA_REVISION_REQUIRED", "Publish a new immutable revision; existing and revoked references cannot be overwritten.")
 _register("DATA_QUOTA_EXCEEDED DATA_QUOTA_UNCONFIGURED", "Ask a project administrator to configure publication storage quota. Revoking publications does not free retained content.")
 _register("DATA_UPLOAD_CONFLICT DATA_UPLOAD_INCOMPLETE DATA_UPLOAD_RATE_LIMITED", "Inspect the upload with data upload get; resume matching parts or explicitly abort it.")
@@ -182,7 +182,7 @@ TASK_REQUEST_CONFLICT TEMPORARY_DEPENDENCY_REQUIRES_RETAIN TEMPORARY_RESOURCE_RE
 TEMPORARY_SCOPE_CONFLICT TEMPORARY_SCOPE_EXPIRED TRIAL_IDEMPOTENCY_CONFLICT TRIAL_VERSION_CONFLICT USAGE_EVENT_CONFLICT
 PROFILE_CORRUPT ADAPTER_STORAGE_UNAVAILABLE INSUFFICIENT_BALANCE METHOD_NOT_ALLOWED REQUEST_TOO_LARGE
 BATCH_POLICY_NOT_APPROVED BATCH_PROVIDER_UNAVAILABLE CHECKPOINT_ACCESS_DENIED CHECKPOINT_DOWNLOAD_UNAVAILABLE
-CHECKPOINT_MANAGEMENT_UNAVAILABLE CHECKPOINT_MANAGEMENT_UNSUPPORTED CHECKPOINT_NOT_DURABLE CHECKPOINT_RUN_REQUIRED
+CHECKPOINT_MANAGEMENT_UNAVAILABLE CHECKPOINT_MANAGEMENT_UNSUPPORTED CHECKPOINT_NOT_DURABLE CHECKPOINT_TRIAL_REQUIRED
 CHECKPOINT_SIGNER_UNAVAILABLE INVALID_CHECKPOINT_CURSOR INVALID_CHECKPOINT_TTL INVALID_CHECKPOINT_VISIBILITY
 INVALID_EXECUTION_CONFIG INVALID_METRICS_RANGE INVALID_TRAIN_JOB_CURSOR INVALID_TRAIN_JOB_LIST_LIMIT
 INVALID_TRAIN_JOB_METADATA INVALID_TRAIN_JOB_SPEC INVALID_TRAIN_JOB_STATUS LOG_RETENTION_EXPIRED
@@ -329,7 +329,11 @@ def upload_recovery(error: BaseException) -> str:
     if isinstance(upload_id, str) and re.fullmatch(r'upload_[a-f0-9]{32}', upload_id):
         lines.extend([f'  Upload retained: {upload_id}', f'  Inspect with: el data upload get {upload_id}'])
     generated = getattr(error, 'generated_idempotency_key', None)
-    if isinstance(generated, str) and re.fullmatch(r'[a-f0-9]{32}', generated):
+    # A definite refusal (permission, validation, not found) fails the same way again; offer the key only
+    # when the outcome is unknown or retryable.
+    status = getattr(error, 'status_code', None)
+    retryable = not isinstance(status, int) or status >= 500 or status in {408, 429}
+    if retryable and isinstance(generated, str) and re.fullmatch(r'[a-f0-9]{32}', generated):
         lines.append(f'  Retry the same input with --idempotency-key {generated}')
     return '\n'.join(lines)
 

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/cli_output.py

@@ -38,7 +38,8 @@ def table(headers: Sequence[object], rows: Sequence[Sequence[object]]) -> str:
 
 
 def timestamp(value: object) -> str:
-    return compact_time(value, local=True)
+    """Table and summary times: UTC to seconds, labelled; --json keeps the original value."""
+    return compact_time(value, utc=True)
 
 
 def next_page(args: argparse.Namespace, cursor: object, program: str) -> str:

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/lifecycle_cli.py

@@ -18,7 +18,9 @@ def add_management_commands(settings: "argparse._SubParsersAction[argparse.Argum
         if action == 'wait':
             command.add_argument('--timeout-seconds', type=float, default=600)
     cleanup = settings.add_parser('cleanup', help='Clean an explicit set of Project resource IDs')
-    cleanup.add_argument('--file', type=Path, required=True, help='JSON array of resource kind and ID objects')
+    cleanup.add_argument('--file', type=Path, required=True,
+                         help='JSON array of {"kind": KIND, "id": ID} objects; KIND is job (a Train Job), endpoint, adapter '
+                              'or checkpoint, for example [{"kind": "job", "id": "job_..."}]')
     cleanup.add_argument('--request-id', required=True)
     wait_options(cleanup)
     scopes = settings.add_parser('resource-scope', help='Manage temporary resources with server TTL').add_subparsers(dest='scope_action', required=True)

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/envloop/terminal.py

@@ -6,7 +6,7 @@ import sys
 import unicodedata
 from contextlib import contextmanager
 from contextvars import ContextVar
-from datetime import datetime
+from datetime import datetime, timezone
 from typing import Any, TextIO
 
 _NO_COLOR = ContextVar("envloop_cli_no_color", default=False)
@@ -47,15 +47,18 @@ def status_text(text: str) -> str:
     return style(text, code) if code else text
 
 
-def compact_time(value: Any, *, local: bool = False) -> Any:
-    """Render aware times to seconds; preserve unknown or timezone-free values."""
+def compact_time(value: Any, *, utc: bool = False) -> Any:
+    """Render aware times to seconds; preserve unknown or timezone-free values.
+
+    ``utc`` converts to UTC and labels it, so a table read on any machine means the same instant.
+    """
     if not isinstance(value, str):
         return value
     try:
         parsed = datetime.fromisoformat(value.replace("Z", "+00:00"))
         if parsed.tzinfo is not None:
-            if local:
-                parsed = parsed.astimezone()
+            if utc:
+                return parsed.astimezone(timezone.utc).strftime("%Y-%m-%d %H:%M:%S UTC")
             return parsed.strftime("%Y-%m-%d %H:%M:%S")
     except ValueError:
         pass

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/cli_output.py

@@ -38,9 +38,10 @@ def list_layout(items: list[dict[str, Any]], args: argparse.Namespace) -> tuple[
         render = dataset_row
         empty = 'No datasets found.'
     elif args.group == 'checkpoint':
-        headers = ('CHECKPOINT ID', 'TYPE', 'MODEL', 'JOB ID', 'URI')
-        render = lambda item: (item.get('checkpoint_id'), item.get('checkpoint_type'), item.get('base_model'),
-                               item.get('job_id'), item.get('uri'))
+        # The checkpoint ID is the resource_id (cp_…); the save name (adapter-final) repeats across Jobs.
+        headers = ('CHECKPOINT ID', 'NAME', 'TYPE', 'MODEL', 'JOB ID', 'URI')
+        render = lambda item: (item.get('resource_id'), item.get('checkpoint_id'), item.get('checkpoint_type'),
+                               item.get('base_model'), item.get('job_id'), item.get('uri'))
         empty = 'No checkpoints found.'
     elif args.group == "profile" and args.name is None:
         headers = ("NAME", "VERSIONS", "DEFAULT", "UPDATED AT")
@@ -64,7 +65,7 @@ def pagination(command: list[str], args: argparse.Namespace, cursor: object) ->
     command.extend(['--limit', str(args.limit), '--cursor=' + str(cursor)])
     fields = {'dataset': ('type', 'query'),
               'data': ('type', 'ref', 'job_id', 'trial_id'),
-              'checkpoint': ('job_id', 'run_id'), 'profile': ('name',)}
+              'checkpoint': ('job_id', 'trial_id'), 'profile': ('name',)}
     for field in fields.get(args.group, ()):
         if getattr(args, field, None) is not None:
             if field == 'query':

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/client.py

@@ -194,7 +194,7 @@ class Runners:
                  source_ref_kind: str = 'branch', dry_run: bool = False):
         """Publish a Runner package directory as the immutable revision NAME@REVISION.
 
-        This is `client.data.create(kind='runner', ...)` with the native manifest built
+        This is `client.data.create(type='runner', ...)` with the native manifest built
         from `directory` (runner.toml plus its main.py or main.sh entrypoint): the
         platform stores the bytes, derives the package and index SHA-256, creates the
         Runner asset on first use and registers the revision. `revision` is the full
@@ -219,7 +219,7 @@ class Runners:
             raise RemoteError('RUNNER_SOURCE_REF_INVALID')
         name, manifest, files, executables = runner_package(directory, name=name, executables=executables)
         ref = f'envloop://runners/{name}@{revision}'
-        publication = Data(self._server).create(kind='runner', ref=ref, manifest=manifest, files=files,
+        publication = Data(self._server).create(type='runner', ref=ref, manifest=manifest, files=files,
                                                 executables=executables, dry_run=dry_run)
         if dry_run:
             return publication

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/data.py

@@ -114,10 +114,10 @@ class Data:
             raise ValueError('limit must be between 1 and 100')
         return path + '?' + urlencode({key: value for key, value in params.items() if value is not None})
 
-    def list(self, *, kind: Literal['dataset', 'artifact', 'checkpoint', 'runner', 'recipe'], ref: str | None = None, job_id: str | None = None, trial_id: str | None = None, limit: int = 25, cursor: str | None = None) -> dict[str, Any]:
-        if kind not in ('dataset', 'artifact', 'checkpoint', 'runner', 'recipe') or (kind != 'artifact' and (job_id or trial_id)):
-            raise ValueError('select a Data kind and appropriate owner filters')
-        return self.transport.request('GET', self._query(self._path(), kind=kind, ref=ref, job_id=job_id, trial_id=trial_id, limit=limit, cursor=cursor))
+    def list(self, *, type: Literal['dataset', 'artifact', 'checkpoint', 'runner', 'recipe'], ref: str | None = None, job_id: str | None = None, trial_id: str | None = None, limit: int = 25, cursor: str | None = None) -> dict[str, Any]:
+        if type not in ('dataset', 'artifact', 'checkpoint', 'runner', 'recipe') or (type != 'artifact' and (job_id or trial_id)):
+            raise ValueError('select a Data type and appropriate owner filters')
+        return self.transport.request('GET', self._query(self._path(), type=type, ref=ref, job_id=job_id, trial_id=trial_id, limit=limit, cursor=cursor))
 
     def get(self, data_id: str) -> dict[str, Any]:
         return self.transport.request('GET', self._path(data_id))
@@ -131,16 +131,16 @@ class Data:
         from envhub.ans.proto import ANSError, DatasetRef, InvalidManifest, parse_uri
 
         reference = parse_uri(uri)
-        kinds = {'datasets':'dataset', 'artifacts':'artifact', 'model':'checkpoint',
+        data_types = {'datasets':'dataset', 'artifacts':'artifact', 'model':'checkpoint',
                  'runners':'runner', 'recipes':'recipe'}
-        if reference.kind not in kinds or reference.uri != uri:
+        if reference.kind not in data_types or reference.uri != uri:
             raise ValueError('select a complete resource ANS URI')
         if isinstance(reference, DatasetRef) and (not isinstance(split, str) or not split):
             raise ValueError('Dataset download requires a split')
         if not isinstance(reference, DatasetRef) and split is not None:
             raise ValueError('split is only supported for Dataset')
         endpoint = self._path() + '/download'
-        parameters = {'kind': kinds[reference.kind], 'ref': uri, 'split': split}
+        parameters = {'type': data_types[reference.kind], 'ref': uri, 'split': split}
         plan = self.transport.request('GET', self._query(endpoint, **parameters))
         from envhub.ans.client import validate_download_plan
         try:
@@ -179,36 +179,36 @@ class Data:
         return {'ref': uri, 'split': split, 'content_digest': plan['content_digest'],
                 'directory': str(target), 'verified': True}
 
-    def create(self, *, kind: str, ref: str | None = None, manifest: dict[str, Any] | None = None, files: dict[str, Any] | None = None, model: str | None = None, revision: str | None = None,
+    def create(self, *, type: str, ref: str | None = None, manifest: dict[str, Any] | None = None, files: dict[str, Any] | None = None, model: str | None = None, revision: str | None = None,
                executables: Sequence[str] = (), dry_run: bool = False, idempotency_key: str | None = None):
         """Register native content or frozen public-HF metadata in this project."""
         from envloop_data.data_publications import hf_source
 
-        source = hf_source(kind, model, revision, ref=ref, manifest=manifest, files=files,
+        source = hf_source(type, model, revision, ref=ref, manifest=manifest, files=files,
                            executables=executables, dry_run=dry_run)
         if source is not None:
             result = self.transport.request('POST', self._path(), source)
-            return result if dry_run else self._native_result(result, kind)
-        return self._publish(kind, ref, manifest, files, dry_run,
+            return result if dry_run else self._native_result(result, type)
+        return self._publish(type, ref, manifest, files, dry_run,
                              executables=executables, idempotency_key=idempotency_key)
 
     def update(self, data_id: str, *, ref: str | None = None, manifest: dict[str, Any] | None = None, files: dict[str, Any] | None = None, model: str | None = None, revision: str | None = None,
-               executables: Sequence[str] = (), kind: str | None = None, dry_run: bool = False, idempotency_key: str | None = None):
+               executables: Sequence[str] = (), type: str | None = None, dry_run: bool = False, idempotency_key: str | None = None):
         """Publish a new revision of the same asset, preserving its previous revision."""
         from envloop_data.data_publications import hf_source
 
-        source = hf_source('checkpoint' if kind is None else kind, model, revision, ref=ref, manifest=manifest, files=files,
+        source = hf_source('checkpoint' if type is None else type, model, revision, ref=ref, manifest=manifest, files=files,
                            executables=executables, dry_run=dry_run)
         resource = self.get(data_id)
-        if kind is not None and kind != resource['kind']:
-            raise ValueError('update must preserve the publication kind')
+        if type is not None and type != resource['type']:
+            raise ValueError('update must preserve the publication type')
         if source is not None:
-            if (resource['kind'] != 'checkpoint' or resource.get('metadata_only') is not True
+            if (resource['type'] != 'checkpoint' or resource.get('metadata_only') is not True
                     or resource.get('name') != model or not resource.get('model_uri')):
                 raise ValueError('HF update must preserve the native model identity')
             result = self.transport.request('POST', self._path(data_id) + '/revisions', source)
-            return result if dry_run else self._native_result(result, resource['kind'])
-        return self._publish(resource['kind'], ref, manifest, files, dry_run,
+            return result if dry_run else self._native_result(result, resource['type'])
+        return self._publish(resource['type'], ref, manifest, files, dry_run,
                              parent_id=data_id, executables=executables, idempotency_key=idempotency_key)
 
     def delete(self, data_id: str, *, dry_run: bool = False):
@@ -217,9 +217,9 @@ class Data:
         return self.transport.request('DELETE', self._path(data_id) + ('?dry_run=true' if dry_run else ''))
 
     @staticmethod
-    def _native_result(result: Any, kind: str):
+    def _native_result(result: Any, data_type: str):
         resource = result.get('resource', result) if isinstance(result, dict) else None
-        if (isinstance(resource, dict) and resource.get('kind') == kind
+        if (isinstance(resource, dict) and resource.get('type') == data_type
                 and resource.get('native_registered') is True and resource.get('native_ready') is True
                 and isinstance(resource.get('data_id', resource.get('id')), str)
                 and re.fullmatch(r'ans_[A-Za-z0-9_-]+', resource.get('data_id', resource.get('id')))):
@@ -231,37 +231,37 @@ class Data:
                 error.staging_id = staging
         raise error
 
-    def _publish(self, kind: str, ref: str | None, manifest: dict[str, Any] | None, files: dict[str, Any] | None, dry_run: bool, *, parent_id: str | None = None, executables: Sequence[str] = (), idempotency_key: str | None = None):
+    def _publish(self, data_type: str, ref: str | None, manifest: dict[str, Any] | None, files: dict[str, Any] | None, dry_run: bool, *, parent_id: str | None = None, executables: Sequence[str] = (), idempotency_key: str | None = None):
         from envloop_data.data_publications import (
             INLINE_BYTES,
-            KINDS,
+            TYPES,
             executable_files,
             prepare,
         )
 
-        if kind not in KINDS or not isinstance(ref, str) or not ref or type(dry_run) is not bool:
-            raise ValueError('select a publication kind, immutable ANS ref and boolean dry_run')
-        executables = executable_files(kind, manifest, executables)
+        if data_type not in TYPES or not isinstance(ref, str) or not ref or type(dry_run) is not bool:
+            raise ValueError('select a publication type, immutable ANS ref and boolean dry_run')
+        executables = executable_files(data_type, manifest, executables)
         with prepare(manifest, files) as (manifest, entries, payload, total, digest):
             if total <= INLINE_BYTES and len(entries) <= 100:
                 encoded = {entry['path']: base64.b64encode(payload.read(cast(int, entry['size']))).decode() for entry in entries}
-                body: dict[str, Any] = {'kind': kind, 'ref': ref, 'manifest': manifest, 'files': encoded, 'dry_run': dry_run}
+                body: dict[str, Any] = {'type': data_type, 'ref': ref, 'manifest': manifest, 'files': encoded, 'dry_run': dry_run}
                 if executables:
                     body['executables'] = executables
                 path = self._path(parent_id) + '/revisions' if parent_id else self._path()
                 result = self.transport.request('POST', path, body)
-                return result if dry_run else self._native_result(result, kind)
+                return result if dry_run else self._native_result(result, data_type)
             request_key = uuid4().hex if idempotency_key is None else idempotency_key
             identifier = None
             try:
-                upload = self.create_upload(kind=kind, ref=ref, manifest=manifest, payload_sha256=digest,
+                upload = self.create_upload(type=data_type, ref=ref, manifest=manifest, payload_sha256=digest,
                                             idempotency_key=request_key, parent_id=parent_id,
                                             executables=executables, dry_run=dry_run)
                 if dry_run:
                     return upload
                 identifier = upload.get('upload_id')
                 if upload.get('state') == 'committed':
-                    return self._native_result(upload, kind)
+                    return self._native_result(upload, data_type)
                 identifier = upload['upload_id']
                 part_bytes = upload['limits']['part_bytes']
                 if type(part_bytes) is not int or not 0 < part_bytes <= INLINE_BYTES:
@@ -278,7 +278,7 @@ class Data:
                         self.upload_part(identifier, number, chunk)
                 if existing:
                     raise RemoteError('DATA_UPLOAD_CONFLICT')
-                return self._native_result(self.complete_upload(identifier), kind)
+                return self._native_result(self.complete_upload(identifier), data_type)
             except BaseException as error:
                 # Keep server state recoverable; do not silently abort another
                 # process's resumable upload or lose its reservation identifier.
@@ -289,16 +289,16 @@ class Data:
                     cast(Any, error).upload_id = identifier
                 raise
 
-    def create_upload(self, *, kind: str, ref: str | None, manifest: dict[str, Any] | None, payload_sha256: str, idempotency_key: str, parent_id: str | None = None,
+    def create_upload(self, *, type: str, ref: str | None, manifest: dict[str, Any] | None, payload_sha256: str, idempotency_key: str, parent_id: str | None = None,
                       executables: Sequence[str] = (), dry_run: bool = False):
-        from envloop_data.data_publications import KINDS, executable_files
+        from envloop_data.data_publications import TYPES, executable_files
 
-        if (kind not in KINDS or not isinstance(ref, str) or not ref or type(dry_run) is not bool
+        if (type not in TYPES or not isinstance(ref, str) or not ref or not isinstance(dry_run, bool)
                 or not isinstance(payload_sha256, str) or not re.fullmatch(r'[a-f0-9]{64}', payload_sha256)
                 or not isinstance(idempotency_key, str) or not re.fullmatch(r'[A-Za-z0-9_.:-]{1,128}', idempotency_key)):
             raise ValueError('invalid upload metadata or idempotency key')
-        executables = executable_files(kind, manifest, executables)
-        body: dict[str, Any] = dict(kind=kind, ref=ref, manifest=manifest, payload_sha256=payload_sha256, dry_run=dry_run)
+        executables = executable_files(type, manifest, executables)
+        body: dict[str, Any] = dict(type=type, ref=ref, manifest=manifest, payload_sha256=payload_sha256, dry_run=dry_run)
         if executables:
             body['executables'] = executables
         if parent_id is not None:

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/data_cli.py

@@ -10,7 +10,7 @@ from typing import Any
 
 from envloop_data.data_publications import (
     INLINE_BYTES,
-    KINDS,
+    TYPES,
     open_file,
     prepare,
     storage_files,
@@ -20,7 +20,7 @@ from envloop_data.data_publications import (
 def add_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParser]", listing: Callable[[Any], argparse.ArgumentParser]):
     data = groups.add_parser('data', help='Manage Project ANS resources and verified downloads').add_subparsers(dest='action', required=True)
     command = listing(data)
-    command.add_argument('--type', required=True, choices=[*KINDS, 'artifact'])
+    command.add_argument('--type', required=True, choices=[*TYPES, 'artifact'])
     for option in ('ref', 'job-id', 'trial-id'):
         command.add_argument('--' + option)
     for action in ('get', 'files'):
@@ -33,7 +33,7 @@ def add_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParser]",
             command.set_defaults(_list_view=True)
     for action in ('create', 'update'):
         command = data.add_parser(action, help='Register native content or HF metadata' if action == 'create' else 'Register a new immutable revision')
-        command.add_argument('--type', required=action == 'create', choices=KINDS)
+        command.add_argument('--type', required=action == 'create', choices=TYPES)
         if action == 'update':
             command.add_argument('data_id', metavar='DATA_ID')
         publication_arguments(command, source=True)
@@ -51,7 +51,7 @@ def add_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParser]",
     command.add_argument('--json', action='store_true')
     uploads = data.add_parser('upload', help='Explicit resumable upload lifecycle').add_subparsers(dest='upload_action', required=True)
     command = uploads.add_parser('create', help='Verify local input and reserve a multipart upload')
-    command.add_argument('--type', choices=KINDS, required=True)
+    command.add_argument('--type', choices=TYPES, required=True)
     publication_arguments(command)
     command.add_argument('--idempotency-key', required=True)
     command.add_argument('--parent-id', help='Existing publication when creating a new revision')
@@ -83,15 +83,19 @@ def add_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParser]",
     command.add_argument('--json', action='store_true')
 
 
+MANIFEST_HELP = ('Native JSON content manifest; a record Dataset source.repository is https://github.com/OWNER/REPO '
+                 'or https://huggingface.co/datasets/ORG/NAME, with source.commit the full commit')
+
+
 def publication_arguments(command: argparse.ArgumentParser, *, source: bool = False):
     command.add_argument('--ref', required=not source, help='Exact immutable ANS URI')
     if source:
         inputs = command.add_mutually_exclusive_group(required=True)
-        inputs.add_argument('--manifest', type=Path, help='Native JSON content manifest')
+        inputs.add_argument('--manifest', type=Path, help=MANIFEST_HELP)
         inputs.add_argument('--model', help='Public Hugging Face model namespace/name; metadata only')
         command.add_argument('--revision', help='Explicit HF source ref; the server freezes its full commit')
     else:
-        command.add_argument('--manifest', type=Path, required=True, help='Native JSON content manifest')
+        command.add_argument('--manifest', type=Path, required=True, help=MANIFEST_HELP)
     command.add_argument('--directory', type=Path, required=not source, help='Directory containing exactly the declared files')
     command.add_argument('--executable', action='append', dest='executables', default=[],
                          help='Declared Runner file that must be executable; repeat for each path')
@@ -128,7 +132,7 @@ def publication_input(args: argparse.Namespace):
 
 def execute(data: Any, args: argparse.Namespace):
     if args.action == 'list':
-        return data.list(kind=args.type, ref=args.ref, job_id=args.job_id, trial_id=args.trial_id, limit=args.limit, cursor=args.cursor)
+        return data.list(type=args.type, ref=args.ref, job_id=args.job_id, trial_id=args.trial_id, limit=args.limit, cursor=args.cursor)
     if args.action == 'get':
         return {'resource': data.get(args.data_id), 'files': data.files(args.data_id, limit=args.limit, cursor=args.cursor)}
     if args.action == 'files':
@@ -148,8 +152,8 @@ def execute(data: Any, args: argparse.Namespace):
             manifest, files = publication_input(args)
             options.update(ref=args.ref, manifest=manifest, files=files, executables=args.executables)
         if args.action == 'create':
-            return data.create(kind=args.type, **options)
-        return data.update(args.data_id, kind=args.type, **options)
+            return data.create(type=args.type, **options)
+        return data.update(args.data_id, type=args.type, **options)
     if args.action == 'register':
         return register(data, args)
     if args.action == 'registration':
@@ -160,7 +164,7 @@ def execute(data: Any, args: argparse.Namespace):
         if args.upload_action == 'create':
             manifest, files = publication_input(args)
             with prepare(manifest, files) as (manifest, _, _, _, digest):
-                return data.create_upload(kind=args.type, ref=args.ref, manifest=manifest, payload_sha256=digest,
+                return data.create_upload(type=args.type, ref=args.ref, manifest=manifest, payload_sha256=digest,
                                           idempotency_key=args.idempotency_key, parent_id=args.parent_id,
                                           executables=args.executables, dry_run=args.dry_run)
         if args.upload_action == 'part':

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/data_publications.py

@@ -11,7 +11,7 @@ from contextlib import contextmanager
 from pathlib import Path
 from typing import Any
 
-KINDS = ('dataset', 'checkpoint', 'runner', 'recipe')
+TYPES = ('dataset', 'checkpoint', 'runner', 'recipe')
 INLINE_BYTES = 8 * 1024 * 1024
 MAX_FILE = 512 * 1024 * 1024
 MAX_TOTAL = 1024 * 1024 * 1024
@@ -50,23 +50,23 @@ def storage_files(manifest: dict[str, Any]) -> list[dict[str, Any]]:
     return sorted(projected, key=lambda item: item['path'])
 
 
-def executable_files(kind: str, manifest: Any, executables: Sequence[str]) -> list[str]:
+def executable_files(data_type: str, manifest: Any, executables: Sequence[str]) -> list[str]:
     """Explicit Runner inventory; do not modify its immutable native manifest."""
     if not isinstance(executables, (tuple, list)):
         raise ValueError('executables must be a sequence of declared Runner paths')
     entries = storage_files(manifest)
     paths = {entry['path'] for entry in entries}
     if (any(not isinstance(path, str) or path not in paths for path in executables)
-            or len(set(executables)) != len(executables) or executables and kind != 'runner'):
+            or len(set(executables)) != len(executables) or executables and data_type != 'runner'):
         raise ValueError('executables must select distinct declared Runner files')
     return sorted(executables)
 
 
-def hf_source(kind: str, model: str | None, revision: str | None, *, ref: Any, manifest: Any, files: Any, executables: Any,
+def hf_source(data_type: str, model: str | None, revision: str | None, *, ref: Any, manifest: Any, files: Any, executables: Any,
               dry_run: bool):
     if model is None and revision is None:
         return None
-    if (kind != 'checkpoint' or ref is not None or manifest is not None or files is not None
+    if (data_type != 'checkpoint' or ref is not None or manifest is not None or files is not None
             or executables or type(dry_run) is not bool
             or not isinstance(model, str) or len(model) > 193
             or not re.fullmatch(r'[A-Za-z0-9][A-Za-z0-9._-]{0,95}/[A-Za-z0-9][A-Za-z0-9._-]{0,95}', model)
@@ -74,7 +74,7 @@ def hf_source(kind: str, model: str | None, revision: str | None, *, ref: Any, m
             or not re.fullmatch(r'[A-Za-z0-9][A-Za-z0-9._/-]*', revision) or '..' in revision
             or any(part in ('', '.') for part in revision.split('/'))):
         raise ValueError('HF metadata requires checkpoint model and explicit revision, without content inputs')
-    return {'kind': 'checkpoint', 'source': {'type': 'hf', 'model': model, 'revision': revision}, 'dry_run': dry_run}
+    return {'type': 'checkpoint', 'source': {'type': 'hf', 'model': model, 'revision': revision}, 'dry_run': dry_run}
 
 
 @contextmanager

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/data_upload.py

@@ -234,7 +234,7 @@ class Registrations:
             if identity.get('name') != dataset or identity.get('version') != version:
                 raise _invalid(f'dataset.toml declares {identity.get("name")}@{identity.get("version")}, not {dataset}@{version}')
             size, sha256 = _digest(bundle)
-            body = {'kind': 'dataset', 'ref': f'envloop://datasets/{dataset}@{version}', 'payload_sha256': sha256,
+            body = {'type': 'dataset', 'ref': f'envloop://datasets/{dataset}@{version}', 'payload_sha256': sha256,
                     'manifest': {'schema': 'envhub.artifact/v1', 'kind': 'file', 'metadata': BUNDLE,
                                  'files': [{'path': name, 'size': size, 'sha256': sha256}]}}
             workspace, project = self.data._scope()

{envloop_cli-1.28.0 → envloop_cli-1.30.0}/packages/data/envloop_data/navigation_resources.py

@@ -62,7 +62,7 @@ class Datasets:
 
 
 def _checkpoint_matches(checkpoint: str, item: dict[str, Any]) -> bool:
-    if checkpoint in {item.get('checkpoint_id'), item.get('uri'), item.get('model_uri')}:
+    if checkpoint in {item.get('resource_id'), item.get('checkpoint_id'), item.get('uri'), item.get('model_uri')}:
         return True
     parsed = urlsplit(checkpoint)
     if parsed.scheme not in {'envloop', 'tinker'} or parsed.query or parsed.fragment:
@@ -133,13 +133,13 @@ class Checkpoints:
     def delete(self, checkpoint_id: str) -> Any:
         return self.transport.request('DELETE', f"/v1/checkpoints/{quote(checkpoint_id, safe='')}")
 
-    def download(self, checkpoint: str, directory: str | Path, *, job_id: str | None = None, run_id: str | None = None) -> dict[str, Any]:
+    def download(self, checkpoint: str, directory: str | Path, *, job_id: str | None = None, trial_id: str | None = None) -> dict[str, Any]:
         from pathlib import Path
 
         directory = Path(directory)
         matches, cursor, seen = cast(list[dict[str, Any]], []), None, set()
         while True:
-            page = self.list(job_id=job_id, run_id=run_id, cursor=cursor, limit=100)
+            page = self.list(job_id=job_id, trial_id=trial_id, cursor=cursor, limit=100)
             for item in page['items']:
                 if _checkpoint_matches(checkpoint, item):
                     matches.append(item)
@@ -166,14 +166,16 @@ class Checkpoints:
         if urlsplit(item.get('uri', '')).scheme == 'mock' and item.get('validation_mode') == 'mock':
             return self._download_archive(item, files, directory)
         transport = self.transport
+        # The checkpoint ID (resource_id) and a uri are both accepted; sign with the reference the caller gave.
+        reference = {'resource_id': checkpoint} if checkpoint == item.get('resource_id') else {'uri': checkpoint}
 
         class Content:
             urls: dict[str, Any] | None = None
 
             def refresh(self) -> None:
                 # Keep the original authorized identity and file hashes across refreshes.
-                signed = transport.request('GET', query('/v1/checkpoints/download', uri=checkpoint,
-                                                        job_id=item['job_id'], run_id=run_id))
+                signed = transport.request('GET', query('/v1/checkpoints/download', **reference,
+                                                        job_id=item['job_id'], trial_id=item.get('trial_id')))
                 urls = signed.get('urls', {})
                 if (set(names) != set(urls)
                         or signed.get('checkpoint_id', item['checkpoint_id']) != item['checkpoint_id']):